Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A system comprising: one or more processors; and a non-transitory computer readable medium comprising instructions that, when executed by the one or more processors, cause the system to perform operations comprising: receiving sensor data; determining an object in an environment represented in the sensor data; inputting at least a portion of the sensor data into a machine learning algorithm; receiving, based at least in part on the portion of the sensor data and from the machine learning algorithm, output associated with a physical parameter of the object, wherein the machine learning algorithm comprises: a coarse output branch configured to output a coarse output; and a fine offset branch configured to output an offset with respect to the coarse output by the coarse output branch; and wherein the output comprises a sum of the offset and a highest confidence value of a set of confidence values associated with the coarse output.
A system processes sensor data to identify an object and estimate its physical parameters. It inputs a portion of this data into a machine learning algorithm. This algorithm has two branches: a coarse output branch that produces an initial estimate represented by a set of confidence values, and a fine offset branch that calculates an adjustment. The system's final estimate for the object's physical parameter is determined by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
2. The system of claim 1 , wherein: a confidence value of the set of confidence values is associated with a potential physical parameter associated with the object.
This system processes sensor data to identify an object and estimate its physical parameters. It inputs a portion of this data into a machine learning algorithm. This algorithm has two branches: a coarse output branch that produces an initial estimate as a set of confidence values, where each confidence value represents a potential physical parameter for the object. A fine offset branch calculates an adjustment. The system's final estimate for the object's physical parameter is determined by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
3. The system of claim 1 , the operations further comprising determining, based at least in part on the sensor data, a two dimensional bounding box associated with the object, wherein: the sensor data comprises image data, the inputting is based at least in part on the two dimensional bounding box; the output associated with the physical parameter of the object comprises: an orientation of a three dimensional bounding box associated with the object; and dimensions of the three dimensional bounding box; and the coarse output represents a coarse orientation of the three dimensional bounding box; and the offset represents an orientation offset with respect to the coarse orientation of the three dimensional bounding box.
A system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It then inputs the portion of the image data within this two-dimensional bounding box into a machine learning algorithm. This algorithm estimates physical parameters of the object, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm has a coarse output branch that provides a coarse estimate of the 3D bounding box's orientation, represented by a set of confidence values. A fine offset branch calculates an orientation offset relative to this coarse orientation. The system's final estimated 3D bounding box orientation is derived by summing this fine offset with the highest confidence value from the coarse output branch's orientation estimate.
4. The system of claim 3 , wherein: the orientation of the three dimensional bounding box is based at least in part on the coarse orientation and the orientation offset, the orientation represented as an angle between: a first ray originating from a center of a sensor associated with the sensor data and passing through a center of the two dimensional bounding box, and a second ray aligned with a direction of the object.
This system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It inputs the image data within this 2D bounding box into a machine learning algorithm to estimate the object's three-dimensional bounding box orientation and dimensions. The algorithm uses a coarse output branch for an initial, coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is calculated by combining this offset with the highest confidence coarse orientation. This orientation is specifically defined as the angle between two rays: a first ray originating from the sensor's center and passing through the center of the 2D bounding box, and a second ray aligned with the object's direction.
5. The system of claim 3 , the operations further comprising estimating a position of the three dimensional bounding box by associating the three dimensional bounding box with the sensor data.
A system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It then inputs the portion of the image data within this two-dimensional bounding box into a machine learning algorithm. This algorithm estimates physical parameters of the object, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm has a coarse output branch that provides a coarse estimate of the 3D bounding box's orientation (as confidence values) and a fine offset branch that calculates an orientation offset relative to this coarse orientation. The system's final estimated 3D bounding box orientation is derived by summing this fine offset with the highest confidence coarse orientation. Additionally, the system estimates the position of the three-dimensional bounding box by correlating it with the original sensor data.
6. The system of claim 5 , wherein: estimating the position of the three dimensional bounding box in the environment comprises minimizing a difference between an association of the three dimensional bounding box with the image data and the two dimensional bounding box.
This system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It then inputs the image data within this 2D bounding box into a machine learning algorithm. This algorithm estimates physical parameters of the object, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm utilizes a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is derived by summing this fine offset with the highest confidence coarse orientation. The system also estimates the 3D bounding box's position in the environment by minimizing the difference between how the 3D bounding box projects onto the image data and the previously detected 2D bounding box.
7. The system of claim 3 , wherein the machine learning algorithm is a convolution neural network trained based at least in part on training data comprising a training two dimensional bounding box and an associated ground truth three dimensional bounding box.
A system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It then inputs the portion of the image data within this two-dimensional bounding box into a machine learning algorithm, which is a convolutional neural network. This CNN estimates physical parameters of the object, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm has a coarse output branch that provides a coarse estimate of the 3D bounding box's orientation (as confidence values) and a fine offset branch that calculates an orientation offset relative to this coarse orientation. The system's final estimated 3D bounding box orientation is derived by summing this fine offset with the highest confidence coarse orientation. The convolutional neural network was trained using data that included training 2D bounding boxes and their corresponding ground truth 3D bounding boxes.
8. The system of claim 7 , wherein: the training data is based at least in part on a transformation to a training image; and the transformation comprises at least one of: mirroring the training image; adding noise to the training image; resizing the training image; or resizing the training two dimensional bounding box.
This system processes image data from sensors to identify an object and determine a two-dimensional bounding box around it. It then inputs the image data within this 2D bounding box into a convolutional neural network (CNN) to estimate the object's three-dimensional bounding box orientation and dimensions. The CNN has a coarse output branch for an initial, coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is calculated by combining this offset with the highest confidence coarse orientation. The CNN was trained using data that included training 2D bounding boxes and their corresponding ground truth 3D bounding boxes. This training data was augmented by transformations applied to training images, such as mirroring, adding noise, resizing the image, or resizing the training 2D bounding box.
9. A method comprising: receiving sensor data; determining an object in an environment represented in the sensor data; inputting at least a portion of the sensor data into a machine learning algorithm; receiving, based at least in part on the portion of the sensor data and from the machine learning algorithm, output associated with a physical parameter of the object, wherein the machine learning algorithm comprises: a coarse output branch configured to output a coarse output; and a fine offset branch configured to output an offset with respect to the coarse output by the coarse output branch; and wherein the output comprises a sum of the offset and a highest confidence value of a set of confidence values associated with the coarse output.
A method for estimating an object's physical parameters involves receiving sensor data and identifying an object within that data. A portion of the sensor data is then fed into a machine learning algorithm. This algorithm uses a coarse output branch to generate an initial estimate, represented by a set of confidence values, and a fine offset branch to produce an adjustment. The method determines the final physical parameter output by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
10. The method of claim 9 , wherein: a confidence value of the set of confidence values is associated with a potential physical parameter associated with the object.
A method for estimating an object's physical parameters involves receiving sensor data and identifying an object within that data. A portion of the sensor data is then fed into a machine learning algorithm. This algorithm uses a coarse output branch to generate an initial estimate as a set of confidence values, where each confidence value is associated with a potential physical parameter for the object. A fine offset branch produces an adjustment. The method determines the final physical parameter output by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
11. The method of claim 9 , further comprising: determining, based at least in part on the sensor data, a two dimensional bounding box associated with the object, wherein: the sensor data comprises image data, the inputting is based at least in part on the two dimensional bounding box; the output associated with the physical parameter of the object comprises: an orientation of a three dimensional bounding box associated with the object; and dimensions of the three dimensional bounding box; and the coarse output represents a coarse orientation of the three dimensional bounding box; and the offset represents an orientation offset with respect to the coarse orientation of the three dimensional bounding box.
A method for estimating physical parameters of an object involves receiving image data from sensors, then determining a two-dimensional bounding box associated with the object in that data. A portion of the image data within this 2D bounding box is then inputted into a machine learning algorithm. This algorithm outputs physical parameters, specifically the orientation and dimensions of a three-dimensional bounding box for the object. The algorithm consists of a coarse output branch, which generates a coarse orientation estimate for the 3D bounding box (represented by confidence values), and a fine offset branch, which generates an orientation offset relative to the coarse orientation. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation.
12. The method of claim 11 , wherein the orientation of the three dimensional bounding box is based at least in part on the coarse orientation and the orientation offset, the orientation represented as an angle between: a first ray originating from a center of a sensor associated with the sensor data and passing through a center of the two dimensional bounding box, and a second ray aligned with a direction of the object.
A method for estimating physical parameters of an object involves receiving image data from sensors, then determining a two-dimensional bounding box for the object. The image data within this 2D bounding box is inputted into a machine learning algorithm, which outputs the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is calculated by combining this offset with the highest confidence coarse orientation. This orientation is specifically defined as the angle between a first ray originating from the sensor's center and passing through the 2D bounding box's center, and a second ray aligned with the object's direction.
13. The method of claim 11 , further comprising: estimating a position of the three dimensional bounding box by associating the three dimensional bounding box with the sensor data.
A method for estimating physical parameters of an object involves receiving image data from sensors, then determining a two-dimensional bounding box associated with the object. The image data within this 2D bounding box is inputted into a machine learning algorithm to output the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation. Additionally, the method includes estimating the position of the three-dimensional bounding box by associating it with the original sensor data.
14. The method of claim 13 , wherein estimating the position of the three dimensional bounding box in the environment comprises minimizing a difference between an association of the three dimensional bounding box with the image data and the two dimensional bounding box.
A method for estimating physical parameters of an object involves receiving image data from sensors, then determining a two-dimensional bounding box for the object. The image data within this 2D bounding box is inputted into a machine learning algorithm to output the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation. The method further estimates the 3D bounding box's position in the environment by minimizing the difference between how the 3D bounding box projects onto the image data and the previously detected 2D bounding box.
15. A non-transitory computer readable medium comprising instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising: receiving sensor data; determining an object in an environment represented in the sensor data; inputting at least a portion of the sensor data into a machine learning algorithm; receiving, based at least in part on the portion of the sensor data and from the machine learning algorithm, output associated with a physical parameter of the object, wherein the machine learning algorithm comprises: a coarse output branch configured to output a coarse output; and a fine offset branch configured to output an offset with respect to the coarse output by the coarse output branch; and wherein the output comprises a sum of the offset and a highest confidence value of a set of confidence values associated with the coarse output.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate an object's physical parameters. The system receives sensor data, identifies an object within it, and inputs a portion of this data into a machine learning algorithm. This algorithm has a coarse output branch producing an initial estimate (as a set of confidence values) and a fine offset branch calculating an adjustment. The instructions cause the system to determine the final physical parameter output by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
16. The non-transitory computer readable medium of claim 15 , wherein: a confidence value of the set of confidence values is associated with a potential physical parameter associated with the object.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate an object's physical parameters. The system receives sensor data, identifies an object within it, and inputs a portion of this data into a machine learning algorithm. This algorithm has a coarse output branch producing an initial estimate as a set of confidence values, where each confidence value is associated with a potential physical parameter for the object. A fine offset branch calculates an adjustment. The instructions cause the system to determine the final physical parameter output by summing this fine offset with the highest confidence value from the coarse output branch's initial estimate.
17. The non-transitory computer readable medium of claim 15 , the operations further comprising: determining, based at least in part on the sensor data, a two dimensional bounding box associated with the object, wherein: the sensor data comprises image data, the inputting is based at least in part on the two dimensional bounding box; the output associated with the physical parameter of the object comprises: an orientation of a three dimensional bounding box associated with the object; and dimensions of the three dimensional bounding box; and the coarse output represents a coarse orientation of the three dimensional bounding box; and the offset represents an orientation offset with respect to the coarse orientation of the three dimensional bounding box.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate physical parameters of an object. The system receives image data from sensors, determines a two-dimensional bounding box for the object, and inputs the image data within this 2D bounding box into a machine learning algorithm. This algorithm outputs physical parameters, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm comprises a coarse output branch that provides a coarse orientation estimate for the 3D bounding box (represented by confidence values), and a fine offset branch that calculates an orientation offset relative to the coarse orientation. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation.
18. The non-transitory computer readable medium of claim 17 , wherein the orientation of the three dimensional bounding box is based at least in part on the coarse orientation and the orientation offset, the orientation represented as an angle between: a first ray originating from a center of a sensor associated with the sensor data and passing through a center of the two dimensional bounding box, and a second ray aligned with a direction of the object.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate physical parameters of an object. The system receives image data from sensors, determines a two-dimensional bounding box for the object, and inputs the image data within this 2D bounding box into a machine learning algorithm. This algorithm outputs the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is calculated by combining this offset with the highest confidence coarse orientation. This orientation is specifically defined as the angle between a first ray originating from the sensor's center and passing through the 2D bounding box's center, and a second ray aligned with the object's direction.
19. The non-transitory computer readable medium of claim 17 , the operations further comprising: estimating a position of the three dimensional bounding box by associating the three dimensional bounding box with the sensor data.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate physical parameters of an object. The system receives image data from sensors, determines a two-dimensional bounding box for the object, and inputs the image data within this 2D bounding box into a machine learning algorithm. This algorithm outputs physical parameters, specifically the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation. Additionally, the instructions cause the system to estimate the position of the three-dimensional bounding box by associating it with the original sensor data.
20. The non-transitory computer readable medium of claim 19 , wherein estimating the position of the three dimensional bounding box in the environment comprises minimizing a difference between an association of the three dimensional bounding box with the image data and the two dimensional bounding box.
A non-transitory computer readable medium stores instructions that, when executed, enable a system to estimate physical parameters of an object. The system receives image data from sensors, determines a two-dimensional bounding box for the object, and inputs the image data within this 2D bounding box into a machine learning algorithm. This algorithm outputs the orientation and dimensions of a three-dimensional bounding box. The algorithm uses a coarse output branch for a coarse orientation estimate (as confidence values) and a fine offset branch for an orientation offset. The final 3D bounding box orientation is determined by summing this offset with the highest confidence coarse orientation. The instructions further cause the system to estimate the 3D bounding box's position in the environment by minimizing the difference between how the 3D bounding box projects onto the image data and the previously detected 2D bounding box.
Unknown
August 4, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.